Abstract: A very relevant problem in e-government is that a great amount of knowledge is in natural language unstructured documents. If that knowledge was stored using a computer-processable representation it would be more easily accessed. In this paper we present the architecture, modules and initial results of a prototype under development for extracting information from government documents. The prototype stores the information using a formal representation of the set of concepts and the relationships between those concepts - an ontology. The system was tested using minutes of Portuguese Municipal Boards meetings. Initial results are presented for an important and frequent topic of the minutes: the subsidies granted by municipalities.
Index Terms: entities and relations extraction, e-government, semantic query.